
from IPython.display import Image
Image(filename=r"C:\Users\Dalkeith J Thomas\OneDrive - The University of the West Indies, Mona Campus\Data Science Projects\Title Page.png")


Image(filename=r"C:\Users\Dalkeith J Thomas\OneDrive - The University of the West Indies, Mona Campus\Data Science Projects\2.png")


Image(filename=r"C:\Users\Dalkeith J Thomas\OneDrive - The University of the West Indies, Mona Campus\Data Science Projects\Aim.png")


import pandas as pd
import warnings
warnings.filterwarnings("ignore")
import matplotlib.pyplot as plt
import seaborn as sns
from scipy.stats import chi2_contingency
from sklearn.model_selection import train_test_split


Image(filename=r"C:\Users\Dalkeith J Thomas\Downloads\Importing and Cleaning.png")


df=pd.read_csv(r"C:\Users\Dalkeith J Thomas\OneDrive - The University of the West Indies, Mona Campus\Data Science Projects\marketing_new.csv")


df.shape

(10037, 16)


df.head()


df=df.drop(columns=['Unnamed: 0','is_correct_lang','channel_code'])


df.head()


df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 10037 entries, 0 to 10036
Data columns (total 13 columns):
 #   Column               Non-Null Count  Dtype  
---  ------               --------------  -----  
 0   user_id              10037 non-null  object 
 1   date_served          10021 non-null  object 
 2   marketing_channel    10022 non-null  object 
 3   variant              10037 non-null  object 
 4   converted            10037 non-null  bool   
 5   language_displayed   10037 non-null  object 
 6   language_preferred   10037 non-null  object 
 7   age_group            10037 non-null  object 
 8   date_subscribed      1856 non-null   object 
 9   date_canceled        577 non-null    object 
 10  subscribing_channel  1856 non-null   object 
 11  is_retained          10037 non-null  bool   
 12  DoW                  1856 non-null   float64
dtypes: bool(2), float64(1), object(10)
memory usage: 882.3+ KB


df_cat=df[["marketing_channel","variant","language_displayed","language_preferred","age_group","subscribing_channel"]]


for col in df_cat:
    print(df[col].value_counts())

House Ads    4733
Instagram    1871
Facebook     1860
Push          993
Email         565
Name: marketing_channel, dtype: int64
control            5091
personalization    4946
Name: variant, dtype: int64
English    9793
Spanish     136
German       81
Arabic       27
Name: language_displayed, dtype: int64
English    9275
Spanish     450
German      167
Arabic      145
Name: language_preferred, dtype: int64
19-24 years    1682
24-30 years    1568
0-18 years     1539
30-36 years    1355
36-45 years    1353
45-55 years    1353
55+ years      1187
Name: age_group, dtype: int64
Instagram    600
Facebook     445
House Ads    354
Email        290
Push         167
Name: subscribing_channel, dtype: int64


df.isna().sum()

user_id                   0
date_served              16
marketing_channel        15
variant                   0
converted                 0
language_displayed        0
language_preferred        0
age_group                 0
date_subscribed        8181
date_canceled          9460
subscribing_channel    8181
is_retained               0
DoW                    8181
dtype: int64


# I realized that Ids were often repeated and subscription date would be filled for one entry
df = df[df['user_id'].notnull() & df['date_subscribed'].notnull()]


df.isna().sum()

user_id                   0
date_served               0
marketing_channel         0
variant                   0
converted                 0
language_displayed        0
language_preferred        0
age_group                 0
date_subscribed           0
date_canceled          1279
subscribing_channel       0
is_retained               0
DoW                       0
dtype: int64


df['is_active'] = df['date_canceled'].isnull().astype('bool')


df.head()


df_date=df[['date_served','date_subscribed','date_canceled',]]


for var in df_date:
    df[var]=pd.to_datetime(df_date[var],format='%Y-%m-%d')


df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1856 entries, 0 to 10036
Data columns (total 14 columns):
 #   Column               Non-Null Count  Dtype         
---  ------               --------------  -----         
 0   user_id              1856 non-null   object        
 1   date_served          1856 non-null   datetime64[ns]
 2   marketing_channel    1856 non-null   object        
 3   variant              1856 non-null   object        
 4   converted            1856 non-null   bool          
 5   language_displayed   1856 non-null   object        
 6   language_preferred   1856 non-null   object        
 7   age_group            1856 non-null   object        
 8   date_subscribed      1856 non-null   datetime64[ns]
 9   date_canceled        577 non-null    datetime64[ns]
 10  subscribing_channel  1856 non-null   object        
 11  is_retained          1856 non-null   bool          
 12  DoW                  1856 non-null   float64       
 13  is_active            1856 non-null   bool          
dtypes: bool(3), datetime64[ns](3), float64(1), object(7)
memory usage: 179.4+ KB


# Create 'Day_of_Week' column
df['DoW'] = df['date_served'].dt.day_name()


df_cat=df[["marketing_channel","variant","language_displayed","language_preferred","age_group","subscribing_channel","DoW"]]


df.head()


Image(filename=r"C:\Users\Dalkeith J Thomas\OneDrive - The University of the West Indies, Mona Campus\Data Science Projects\Exploratory D Analysis.png")


market_channel_cat=df['marketing_channel'].value_counts()

plt.figure(figsize=(8,6))
plt.pie(market_channel_cat, labels=market_channel_cat.index, autopct='%1.1f%%', startangle=140)
plt.title('Pie Chart of Marketing Channel Distribution')
plt.tight_layout()
plt.show()


# Create the bar plot
plt.figure(figsize=(8, 6))  # Set the figure size (width, height) in inches
sns.countplot(x='marketing_channel',data=df,palette="Set3")

# Customize the plot
plt.title('Countplot showing Marketing Channels')  
plt.xlabel('Marketing Channel')  
plt.ylabel('Values')  

# Show the plot
plt.tight_layout() 
plt.show()


for col in df_cat:
    plt.figure(figsize=(10, 6))
    sns.countplot(y=col, data=df, order=df[col].value_counts().index,palette='Set3')
    plt.title(f'Distribution of {col}')
    plt.show()


conversions_over_time = df.groupby(df['date_served'].dt.to_period('D'))['converted'].sum()
conversions_over_time.plot(kind='bar', figsize=(12, 6), title='Conversions Over Time')
plt.xlabel('Day')
plt.ylabel('Number of Conversions')
plt.show()


spike_filtered=['2018-01-15','2018-01-16','2018-01-17']
spike_filtered_df=df[df['date_served'].isin(spike_filtered)]

excluded_df=df[~df['date_served'].isin(spike_filtered)]


for col in df_cat:
    plt.figure(figsize=(10, 6))
    sns.countplot(y=col, data=spike_filtered_df, order=df[col].value_counts().index,palette='Set3')
    plt.title(f'Distribution of {col} for the spike days')
    plt.show()


for col in df_cat:
    plt.figure(figsize=(10, 6))
    sns.countplot(y=col, data=excluded_df, order=df[col].value_counts().index,palette='Set3')
    plt.title(f'Distribution of {col} with exclusions')
    plt.show()


cancel_counts = df['is_retained'].value_counts()


plt.figure(figsize=(8, 6))
plt.pie(cancel_counts, labels=cancel_counts.index, autopct='%1.1f%%', startangle=140)
plt.title('Pie Chart Showing the Percentage of Active Users')

plt.legend(labels=cancel_counts.index, title="User Status", loc="center left", bbox_to_anchor=(1, 0.5))

plt.show()


convert_counts = df['converted'].value_counts()


plt.figure(figsize=(8, 6))
plt.pie(convert_counts, labels=convert_counts.index, autopct='%1.1f%%', startangle=140)
plt.title('Pie Chart Showing the Percentage of Converted Users')

plt.legend(labels=convert_counts.index, title="User Status", loc="center left", bbox_to_anchor=(1, 0.5))

plt.show()


Image(filename=r"C:\Users\Dalkeith J Thomas\OneDrive - The University of the West Indies, Mona Campus\Data Science Projects\Feature Engineering.png")


total_consumers=len(df)
total_converted=df[df['converted']==True].shape[0]
print(f"Total consumers : {total_consumers},Total converted : {total_converted}")

Total consumers : 1856,Total converted : 1050


total_conversion_rate=(total_converted/total_consumers)*100
print(f"The overall conversion rate is {total_conversion_rate:.2f}%")

The overall conversion rate is 56.57%


total_retained=df[df['is_retained']==True].shape[0]
total_retention_rate=(total_retained/total_consumers)*100
print(f"The overall retention rate is {total_retention_rate:.2f}%")

The overall retention rate is 68.91%


rates = {'Conversion Rate': total_conversion_rate, 'Retention Rate': total_retention_rate}
names = list(rates.keys())
values = list(rates.values())
colors = sns.color_palette("Set3", len(names))
plt.figure(figsize=(10, 6))
bars = plt.bar(names, values, color=colors)
plt.title('Conversion and Retention Rates')
plt.ylabel('Rate(%)')
plt.ylim(0, 100) 
for bar in bars.patches:
    y_val=bar.get_height()
    plt.text(bar.get_x()+bar.get_width()/2,y_val+0.05,f'{y_val:.2f}',ha='center',va='bottom')
plt.show()


def conversion(column):
    column_conversion = df.groupby(column)['converted'].mean() * 100
    colors = sns.color_palette('Set3')
    plt.figure(figsize=(10,6))
    bars = column_conversion.plot(kind='bar', color=colors, title=f'Conversion Rate by {column}')
    plt.xlabel(column)
    plt.ylabel('Conversion Rate (%)')
    plt.xticks(rotation=45)
    
    for bar in bars.patches:
        yval = bar.get_height()
        plt.text(bar.get_x() + bar.get_width()/2, yval + 0.5, f'{yval:.2f}%', ha='center', va='bottom')
    
    plt.tight_layout()
    plt.show()


def retention(column):
    column_retention = df.groupby(column)['is_retained'].mean() * 100
    colors = sns.color_palette('Set2')
    plt.figure(figsize=(10,6))
    bars = column_retention.plot(kind='bar', color=colors, title=f'Retention Rate by {column}')
    plt.xlabel(column)
    plt.ylabel('Retention Rate (%)')
    plt.xticks(rotation=45)
    
    for bar in bars.patches:
        yval = bar.get_height()
        plt.text(bar.get_x() + bar.get_width()/2, yval + 0.5, f'{yval:.2f}%', ha='center', va='bottom')
    
    plt.tight_layout()
    plt.show()


for col in df_cat:
    conversion(col)


for col in df_cat:
    retention(col)


def heatmap(var1,var2):
    heatmap_data = pd.crosstab(df[var1], df[var2])
    plt.figure(figsize=(10, 6))
    sns.heatmap(heatmap_data, annot=True, fmt="d", cmap="YlGnBu")
    plt.title(f'Frequency of {var1} by {var2}')
    plt.xlabel(var2)
    plt.ylabel(var1)
    plt.show()


heatmap('age_group','marketing_channel')


def ratemap(val,ind,col):
    heatmap_data = df.pivot_table(values=(val), index=(ind), columns=(col), aggfunc='mean')
    plt.figure(figsize=(10, 6))
    sns.heatmap(heatmap_data, annot=True, fmt=".2f", cmap="viridis")
    plt.title(f'{val} Rates by {col} and {ind}')
    plt.xlabel(col)
    plt.ylabel(ind)
    plt.show()


ratemap('converted','age_group','marketing_channel')


ratemap('converted','age_group','variant')


ratemap('converted','age_group','DoW')


Image(filename=r"C:\Users\Dalkeith J Thomas\OneDrive - The University of the West Indies, Mona Campus\Data Science Projects\AB testing.png")


conversion_rates = df.groupby('variant')['converted'].mean() * 100
print(conversion_rates)

variant
control            37.870472
personalization    74.603175
Name: converted, dtype: float64


heatmap('converted','variant')


contingency_table = pd.crosstab(df['converted'], df['variant'])

chi2, p, dof, ex = chi2_contingency(contingency_table)
print(f"Chi-square: {chi2}, p-value: {p}")

#Assuming Alpha is 0.05
alpha = 0.05  
if p < alpha:
    print("Reject the null hypothesis: There is a significant difference between the groups.")
else:
    print("Fail to reject the null hypothesis: There is no significant difference between the groups.")

Chi-square: 253.25434002346478, p-value: 5.069698946689019e-57
Reject the null hypothesis: There is a significant difference between the groups.


Image(filename=r"C:\Users\Dalkeith J Thomas\OneDrive - The University of the West Indies, Mona Campus\Data Science Projects\Importing and Cleaning.png")


df['converted']=df['converted'].astype(int)
df['is_retained']=df['is_retained'].astype(int)


columns_to_drop = ['user_id', 'date_served','date_subscribed','is_active','date_canceled']
df.drop(columns=columns_to_drop, inplace=True)
encoded_df = pd.get_dummies(df, columns=["marketing_channel","variant","language_displayed","language_preferred","age_group","subscribing_channel","DoW"])

print(encoded_df.head())

   converted  is_retained  marketing_channel_Email  \
0          1            1                        0   
1          1            1                        0   
2          1            1                        0   
3          1            1                        0   
4          1            1                        0   

   marketing_channel_Facebook  marketing_channel_House Ads  \
0                           0                            1   
1                           0                            1   
2                           0                            1   
3                           0                            1   
4                           0                            1   

   marketing_channel_Instagram  marketing_channel_Push  variant_control  \
0                            0                       0                0   
1                            0                       0                0   
2                            0                       0                0   
3                            0                       0                0   
4                            0                       0                0   

   variant_personalization  language_displayed_Arabic  ...  \
0                        1                          0  ...   
1                        1                          0  ...   
2                        1                          0  ...   
3                        1                          0  ...   
4                        1                          0  ...   

   subscribing_channel_House Ads  subscribing_channel_Instagram  \
0                              1                              0   
1                              1                              0   
2                              1                              0   
3                              1                              0   
4                              1                              0   

   subscribing_channel_Push  DoW_Friday  DoW_Monday  DoW_Saturday  DoW_Sunday  \
0                         0           0           1             0           0   
1                         0           0           1             0           0   
2                         0           0           1             0           0   
3                         0           0           1             0           0   
4                         0           0           1             0           0   

   DoW_Thursday  DoW_Tuesday  DoW_Wednesday  
0             0            0              0  
1             0            0              0  
2             0            0              0  
3             0            0              0  
4             0            0              0  

[5 rows x 36 columns]


X = encoded_df.drop('converted', axis=1)
y = encoded_df['converted']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)


from sklearn.preprocessing import MinMaxScaler

scaler = MinMaxScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)


from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier
from sklearn.svm import SVC
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score,confusion_matrix


# Define classifiers
classifiers = {
    'Logistic Regression': LogisticRegression(),
    'Random Forest': RandomForestClassifier(),
    'Support Vector Machine': SVC(probability=True),
    'K-Nearest Neighbors': KNeighborsClassifier(),
    'Decision Tree': DecisionTreeClassifier(),
    'Naive Bayes': GaussianNB()
}


for clf_name, clf in classifiers.items():
    # Train the classifier
    clf.fit(X_train, y_train)
    
    # Make predictions on the testing data
    y_pred = clf.predict(X_test)
    
    # Calculate accuracy, precision, and recall
    accuracy = accuracy_score(y_test, y_pred)
    precision = precision_score(y_test, y_pred)
    recall = recall_score(y_test, y_pred)
    f1=f1_score(y_test,y_pred)
    
     # Calculate confusion matrix
    cm = confusion_matrix(y_test, y_pred)
    
    # Plot confusion matrix
    plt.figure(figsize=(8, 6))
    sns.heatmap(cm, annot=True, cmap='Blues', fmt='g')
    plt.title(f'Confusion Matrix - {clf_name}')
    plt.xlabel('Predicted Labels')
    plt.ylabel('True Labels')
    plt.show()
    
    # Print the scores for the current classifier
    print(f'{clf_name} Metrics:')
    print(f'Accuracy: {accuracy:.4f}')
    print(f'Precision: {precision:.4f}')
    print(f'Recall: {recall:.4f}')
    print(f'F1 Score: {f1}')
    print('\n')

Logistic Regression Metrics:
Accuracy: 0.8441
Precision: 0.8263
Recall: 0.9198
F1 Score: 0.8705357142857143

Random Forest Metrics:
Accuracy: 0.9032
Precision: 0.9151
Recall: 0.9151
F1 Score: 0.9150943396226415

Support Vector Machine Metrics:
Accuracy: 0.8898
Precision: 0.8834
Recall: 0.9292
F1 Score: 0.9057471264367817

K-Nearest Neighbors Metrics:
Accuracy: 0.8710
Precision: 0.8661
Recall: 0.9151
F1 Score: 0.8899082568807338

Decision Tree Metrics:
Accuracy: 0.8871
Precision: 0.9250
Recall: 0.8726
F1 Score: 0.8980582524271845

Naive Bayes Metrics:
Accuracy: 0.6317
Precision: 0.8319
Recall: 0.4434
F1 Score: 0.5784615384615385


from sklearn.metrics import roc_curve, roc_auc_score

plt.figure(figsize=(8, 6))

# Loop through classifiers and plot ROC curves
for clf_name, clf in classifiers.items():
    # Make predictions on the test data to get predicted probabilities
    y_pred_proba = clf.predict_proba(X_test)[:, 1]
    
    # Calculate ROC curve and AUC score
    fpr, tpr, _ = roc_curve(y_test, y_pred_proba)
    auc_score = roc_auc_score(y_test, y_pred_proba)
    
    # Plot ROC curve
    plt.plot(fpr, tpr, label=f'{clf_name} (AUC = {auc_score:.2f})')

# Plot the random guess line (diagonal)
plt.plot([0, 1], [0, 1], linestyle='--', color='gray', label='Random Guess')

# Set labels and title
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver Operating Characteristic (ROC) Curve')
plt.legend()
plt.grid(True)
plt.show()


from sklearn.metrics import classification_report
from sklearn.model_selection import RandomizedSearchCV
import numpy as np
from sklearn.ensemble import RandomForestClassifier

# Reduce the size of the parameter grid
param_grid = {
    'n_estimators': [100, 200], 
    'max_depth': [None, 10],  
    'min_samples_split': [2, 5],  
    'min_samples_leaf': [1, 2]  
}

rf = RandomForestClassifier(random_state=42)
random_search = RandomizedSearchCV(rf, param_distributions=param_grid, n_iter=5, cv=3, scoring='accuracy', n_jobs=-1)

random_search.fit(X_train, y_train)

best_params = random_search.best_params_
best_estimator = random_search.best_estimator_

y_pred = best_estimator.predict(X_test)
print(classification_report(y_test, y_pred))
rf=best_estimator

              precision    recall  f1-score   support

           0       0.89      0.89      0.89       160
           1       0.92      0.92      0.92       212

    accuracy                           0.91       372
   macro avg       0.90      0.90      0.90       372
weighted avg       0.91      0.91      0.91       372


print(rf)

RandomForestClassifier(min_samples_split=5, random_state=42)


from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split

# Extract feature importances
feature_importances = rf.feature_importances_

# Create a DataFrame for better visualization
feature_importances_df = pd.DataFrame({
    'Feature': X_train.columns,
    'Importance': feature_importances
})

# Sort the DataFrame by importance
feature_importances_df = feature_importances_df.sort_values(by='Importance', ascending=False)

# Visualize the feature importance
plt.figure(figsize=(12, 8))
sns.barplot(x='Importance', y='Feature', data=feature_importances_df)
plt.title('Feature Importance in Random Forest')
plt.show()


importance_df = pd.DataFrame({'Feature': X_train.columns, 'Importance': feature_importances})
importance_df = importance_df.sort_values(by='Importance', ascending=False)
print(importance_df)

                          Feature  Importance
25  subscribing_channel_House Ads    0.153402
3     marketing_channel_House Ads    0.146371
6                 variant_control    0.089454
7         variant_personalization    0.078922
26  subscribing_channel_Instagram    0.076525
4     marketing_channel_Instagram    0.050185
24   subscribing_channel_Facebook    0.049337
2      marketing_channel_Facebook    0.036749
1         marketing_channel_Email    0.033639
23      subscribing_channel_Email    0.024058
27       subscribing_channel_Push    0.020620
32                   DoW_Thursday    0.020097
34                  DoW_Wednesday    0.018875
5          marketing_channel_Push    0.018051
0                     is_retained    0.017653
22            age_group_55+ years    0.015020
29                     DoW_Monday    0.014997
17          age_group_19-24 years    0.014192
18          age_group_24-30 years    0.011974
20          age_group_36-45 years    0.011857
21          age_group_45-55 years    0.011037
33                    DoW_Tuesday    0.010719
31                     DoW_Sunday    0.010059
28                     DoW_Friday    0.009808
16           age_group_0-18 years    0.009594
30                   DoW_Saturday    0.009521
19          age_group_30-36 years    0.008535
13     language_preferred_English    0.007139
14      language_preferred_German    0.005747
9      language_displayed_English    0.004525
12      language_preferred_Arabic    0.002944
10      language_displayed_German    0.002831
15     language_preferred_Spanish    0.002760
8       language_displayed_Arabic    0.001411
11     language_displayed_Spanish    0.001392

	Unnamed: 0	user_id	date_served	marketing_channel	variant	converted	language_displayed	language_preferred	age_group	date_subscribed	date_canceled	subscribing_channel	is_retained	channel_code	is_correct_lang
0	0	a100000029	2018-01-01	House Ads	personalization	True	English	English	0-18 years	2018-01-01	NaN	House Ads	True	1.0	Yes
1	1	a100000030	2018-01-01	House Ads	personalization	True	English	English	19-24 years	2018-01-01	NaN	House Ads	True	1.0	Yes
2	2	a100000031	2018-01-01	House Ads	personalization	True	English	English	24-30 years	2018-01-01	NaN	House Ads	True	1.0	Yes
3	3	a100000032	2018-01-01	House Ads	personalization	True	English	English	30-36 years	2018-01-01	NaN	House Ads	True	1.0	Yes
4	4	a100000033	2018-01-01	House Ads	personalization	True	English	English	36-45 years	2018-01-01	NaN	House Ads	True	1.0	Yes